Webinar, Interview, Product Demonstration
AI has got better at hacking—how big a risk is it?
- The timeframe for the interval between vulnerability disclosure and exploitation is projected to shrink to one minute by 2028.
- Future open-source models are expected to acquire hacking capabilities comparable to Mythos, creating significant security concerns.
- A demonstration is planned to illustrate the conversion of a specific OpenBSD operating system flaw into a functional exploit.
- Mythos has identified several thousand vulnerabilities, though full disclosure of all details will be withheld for security reasons.
- Immediate patching is advised for OpenBSD users, as the referenced vulnerability has already been resolved through responsible disclosure.
- The successor model following Opus is anticipated to utilize increased data and compute resources to outperform Opus across all capabilities.
- Publicly available models will retain safety restrictions against generating exploits, unlike unsafe models such as Mythos which execute cybersecurity commands without such safeguards.